Structure selective updating for nonlinear models and radial basis function neural networks

Selective model structure and parameter updating algorithms are introduced for both the online estimation of NARMAX models and training of radial basis function neural networks. Techniques for on-line model modification, which depend on the vector-shift properties of regression variables in linear models, cannot be applied when the model is non-linear. In the present paper new methods for on-line model modification are developed. These methods are based on selectively updating the non-linear model structure and therefore lead to a reduction in computational cost. A real data set is used to demonstrate the performance of the new algorithms. © 1998 John Wiley & Sons, Ltd.